Thanks, between that and the source code I know understand what is happening.
The configure script never checked that the flags are supported by the compiler.
Instead it runs some assembly code to identify the cpu ID and infer the
capabilities
from that information. Actual code using the
Oh, my bad. Here it is.
Am Dienstag, 6. August 2019 07:34:02 UTC+2 schrieb François Bissey:
>
> I meant the config.log of fflas-ffpack not sage’s one.
>
> > On 6/08/2019, at 5:31 PM, Markus Wageringel > wrote:
> >
> >
>
>
--
You received this message because you are subscribed to the
I meant the config.log of fflas-ffpack not sage’s one.
> On 6/08/2019, at 5:31 PM, Markus Wageringel
> wrote:
>
>
--
You received this message because you are subscribed to the Google Groups
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an
Yes I would like to see it. I have looked at what fflas-ffpack does to detect
the stuff but I want a practical output for that case.
> On 6/08/2019, at 10:26 AM, Markus Wageringel
> wrote:
>
> It would have been nice to have the config log from when you had the failure.
>
> Would you still
>
> It would have been nice to have the config log from when you had the
> failure.
>
Would you still like to have it? I could reproduce it I suppose.
--
You received this message because you are subscribed to the Google Groups
"sage-release" group.
To unsubscribe from this group and stop
After inspection of configure they are all brand new options I didn’t know
anything
about. That would have been —disable-{avx512f,avx512dq,avx512vl}. I am still
shocked
it enabled them without your compiler supporting it. The logic must be flawed.
In any case we need to add those to the list of
Thank you for the quick replies. It is the system's default compiler – no
idea why it is so old.
export FFLAS_FFPACK_CONFIGURE="-disable-fma -disable-fma4 -disable-avx
> -disable-avx2"
Hm, that did not work, giving still the same error for fflas-ffpack. I
assume I would have to disable AVX
But fflas-ffpack detection routines pick it up when they shouldn’t. Technically
I see that
as an upstream inbox team problem since they share their detection routines
over the entire
givaro/fflas-ffpack/linbox stack.
You may want to try
export FFLAS_FFPACK_CONFIGURE=“—disable-fma
You're using gcc version 4.8.5, released in 2015. I guess AVX 512
didn't exist yet.
On Mon, Aug 5, 2019 at 11:17 PM Markus Wageringel
wrote:
>
> The package fflas_ffpack-2.4.3 added in #26932 fails to compile for me on
> CentOS 7.6.1810, due to several errors of this form:
>
> g++: error:
